StoryTrans: Non-Parallel Story Author-Style Transfer with Discourse Representations and Content Enhancing

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 14803–14819 July 9-14, 2023 ©2023 Association for Computational Linguistics StoryTrans: Non-Parallel Story Author-Style Transfer with Discourse Representations and Content Enhancing Xuekai Zhu2∗, Jian Guan1∗, Minlie Huang1†and Juan Liu2 1The CoAI group, DCST, Institute for Artificial Intelligence, State Key Lab of Intelligent Technology and Systems, 1Beijing National Research Center for Information Science and Technology, Tsinghua University, Beijing 100084, China 2 Institute of Artificial Intelligence, School of Computer Science, Wuhan University, Wuhan, 430072, China {xuekaizhu,liujuan}@whu.edu.cn, j-guan19@mails.tsinghua.edu.cn, aihuang@tsinghua.edu.cn Abstract Non-parallel text style transfer is an impor- tant task in natural language generation. How- ever, previous studies concentrate on the to- ken or sentence level, such as sentence senti- ment and formality transfer, but neglect long style transfer at the discourse level. Long texts usually involve more complicated author lin- guistic preferences such as discourse structures than sentences. In this paper, we formulate the task of non-parallel story author-style trans- fer, which requires transferring an input story into a specified author style while maintaining source semantics. To tackle this problem, we propose a generation model, named StoryTrans, which leverages discourse representations to capture source content information and trans- fer them to target styles with learnable style embeddings. We use an additional training ob- jective to disentangle stylistic features from the learned discourse representation to prevent the model from degenerating to an auto-encoder. Moreover, to enhance content preservation, we design a mask-and-fill framework to explicitly fuse style-specific keywords of source texts into generation. Furthermore, we constructed new datasets for this task in Chinese and English, respectively. Extensive experiments show that our model outperforms strong baselines in over- all performance of style transfer and content preservation. 1 Introduction Text style transfer aims to endow a text with a different style while keeping its main semantic con- tent unaltered. It has a wide range of applications, such as formality transfer (Jain et al., 2019), senti- ment transfer (Shen et al., 2017) and author-style imitation (Tikhonov and Yamshchikov, 2018). Due to the lack of parallel corpora, recent works mainly focus on unsupervised transfer by self- reconstruction. Current methods proposed to dis- ∗Equal contribution. †Corresponding author 郭翰是古时候一名才子。一个夏日的晚上，他在院中乘凉。忽 然，一阵风起，送来一股沁人心脾的清香，一位少女驾着白云从 天而降，出现在郭翰眼前 … Guo Han was a talented man in ancient times. One summer evening, he was enjoying the cool in the courtyard. Suddenly, a gust of wind brought a refreshing fragrance, and a young girl descended from the sky on a white cloud and appeared in front of Guo Han … 郭翰在夏日的夜,院中花香沁人心脾,一阵清香从身旁飘来,那少女 却是神色自若,一言不发的从天而降。郭翰大惊,眼前白光一闪,身 前纱衣一晃,已被她夺了过去。… On a summer night, the scent of flowers is refreshing in the courtyard，a scent of fragrance floats from Guo Han side. A young girl, with a calm expression, fell from the sky without saying a word. Guo Han was shocked, and already taken away by her with a white light flashing and the gauze flickering. … Source Text: Generated Text for JY Style: Table 1: An example that transfers a vernacular story to the martial arts style of JY generated …

Content truncated to 500 words